feat: close the chord that let a teacher hibernate the room

Ctrl+Alt+Delete opened Plasma's session screen and a teacher hibernated a
classroom computer from it. A hibernated room cannot be recovered by
anybody standing in front of it, which on a Sunday morning is everybody.

logout=false was already set and did not stop it. That restriction
governs Plasma's own actions — what plasmashell draws in a menu or an
applet. The chord is a KGlobalAccel shortcut registered by ksmserver and
it asks D-Bus for the greeter without consulting a Plasma action at all.
A restriction on a menu cannot see a key.

Three layers, because each one alone has a way past it.

The key. kglobalshortcutsrc unbinds every ksmserver session action and
PowerDevil's Sleep, Hibernate, PowerDown and PowerOff, each entry marked
immutable. It is deliberately not in the xdg directory with the other
settings: kglobalacceld opens that file as a SimpleConfig, so it reads
one file in the teacher's home and no system directory at all. Measured —
with it in the cascade, kreadconfig6 reported the chord unbound while the
shortcut server went on answering Ctrl+Alt+Del. classroom-desktop-reset
copies the package's copy into place before the shortcut server starts,
the same way the panel is copied out of the package at every login.

The policy. A polkit rule refuses login1.suspend* and .hibernate* for the
classroom user and for nobody else. This is the layer that matters:
a key can be unbound and anything reaching logind can still ask.
Refusing here rather than masking sleep.target keeps a technician's own
account working, and is also what tells Plasma to stop drawing the
button — CanSuspend and CanHibernate come back no. Hybrid sleep and
suspend-then-hibernate have no action of their own in systemd 257; both
are authorised through hibernate, so denying it denies all three.

The greeter. Its D-Bus activation is shadowed with /bin/false, the same
mechanism already used for the old notification daemon. Layer one removes
the caller, this removes the callee.

Measured in the lab, from the package, after a reboot: the escape
reproduced first with the shadow removed — the greeter really does start
— and then refused, exited 1, with it back. The chord reports nothing
bound. systemctl suspend as the classroom user is denied. The maintenance
chord and the display chord both still answer, by key code, after the
shortcut file is replaced wholesale. A root process may still suspend,
hibernate, reboot and power off. 182 checks pass across a logout and back.

Screen locking stays disabled, the power button still does nothing, and
reboot and shut down are not newly denied to anybody.

Co-Authored-By: Claude Opus 5 <[email protected]>
This commit is contained in:
ekklesia
2026-09-19 20:50:36 -05:00
co-authored by Claude Opus 5
parent 8db5e8dd9f
commit 23dff0b7b4
8 changed files with 467 additions and 0 deletions
+118
View File
@@ -0,0 +1,118 @@
# Lockdown
What a teacher can and cannot do to the machine, and where each of those
answers lives.
Most of it is KDE's own Kiosk mechanism, in
`/etc/ekklesia-classroom/xdg/kdeglobals`: no shell, no command line, no
lock screen, no editable panel. That file is the first place to look, and
lifting a restriction for maintenance is a one-line edit there followed by
a session restart.
This page is about the part that is *not* in that file: ending or
suspending the session.
## The escape this closed
A teacher pressed **Ctrl+Alt+Delete**, got Plasma's session screen, and
hibernated the classroom computer from it. A hibernated room cannot be
recovered by anybody standing in front of it who does not know where the
power button is, which on a Sunday morning is everybody.
`[KDE Action Restrictions] logout=false` was already set and did not stop
it, and the reason is worth keeping written down: that restriction governs
**Plasma's own actions** — the entries plasmashell draws in a menu or an
applet. Ctrl+Alt+Delete is not one of them. It is a KGlobalAccel shortcut
registered by ksmserver, and pressing it asks D-Bus for the logout greeter
without ever consulting a Plasma action. A restriction on a menu cannot
see a key.
## Three layers
### 1. The key
`/etc/ekklesia-classroom/kglobalshortcutsrc` unbinds every ksmserver
session action and PowerDevil's Sleep, Hibernate, PowerDown and PowerOff
keys, with each entry marked `[$i]` so nothing writes them back.
It is **not** in `/etc/ekklesia-classroom/xdg` with the other settings, and
that is deliberate. `kglobalacceld` opens `kglobalshortcutsrc` as a
KConfig `SimpleConfig`: no cascade, no system directories. Measured — with
the file in the cascade, `kreadconfig6` reported the chord unbound while
the shortcut server went on answering Ctrl+Alt+Del.
So `classroom-desktop-reset` copies the package's copy into
`~/.config/kglobalshortcutsrc` before the session's shortcut server
starts, exactly as it does with the panel. Replacing the whole file costs
nothing: KGlobalAccel puts every shortcut the file does not name back from
each component's own defaults, and the entries the package does name are
immutable, which is what stops it putting those back too.
The room's own chords are registered from their desktop entries and are
unaffected — `converged-check` asserts both of them by key code after the
replace.
### 2. The policy
`/usr/share/polkit-1/rules.d/49-ekklesia-classroom-sleep.rules` refuses
`org.freedesktop.login1.suspend*` and `.hibernate*` **for the `classroom`
user only**.
This is the layer that matters, because a key can be unbound and anything
that reaches logind can still ask. PowerDevil, the greeter, `systemctl
suspend` from the maintenance terminal and any future applet all go
through these actions.
Hybrid sleep and suspend-then-hibernate have no polkit action of their own
in systemd 257: logind authorises both through the hibernate action, so
denying hibernate denies all three. The rule names the other two spellings
anyway, in case a future systemd splits them out.
Refusing here rather than masking `sleep.target` is the point. Masking is
machine-wide and would take suspend away from a technician's account too;
worse, it would be invisible to Plasma, which would go on drawing a
Hibernate button that silently failed. logind asks polkit whether *this
user* may sleep, so a NO is both the enforcement and the supported way to
tell the desktop not to offer it — `CanSuspend()` and `CanHibernate()`
come back `no`.
Nothing here touches power-off or reboot, and nothing here applies to any
other account.
### 3. The greeter
`/usr/local/share/dbus-1/services/org.kde.LogoutPrompt.service` shadows
plasma-workspace's own activation file with `Exec=/bin/false`, the same
mechanism the classroom already uses for the old notification daemon.
Ctrl+Alt+Delete does not draw anything itself; it asks D-Bus to start
`ksmserver-logout-greeter`. Layer 1 removes the caller, this removes the
callee, so any other caller finds nothing to start either.
## What is deliberately still there
- **Screen locking stays disabled.** None of this touches
`kscreenlockerrc`, and `action/lock_screen=false` is unchanged. A
classroom must never present a password prompt.
- **The power button still does nothing** on a short press, at both the
logind and PowerDevil layers, and a long firmware hold still cuts the
power. That is the only recovery a person standing at a wedged machine
has.
- **Meta+Ctrl+Alt+Shift+K still opens the maintenance terminal**, and
Meta+Ctrl+Alt+Shift+D still opens the display dialog.
- **Reboot and shut down are not newly denied to anybody.** A teacher
asking for either already lands on an administrator's password prompt,
which they do not have; that behaviour is unchanged. A technician on SSH
or at the maintenance terminal, with their own account, can still do all
of it.
## Lifting it
For a technician who needs a classroom machine to suspend — a laptop on a
bench, say — drop a file of the same name into `/etc/polkit-1/rules.d/`,
which polkit reads ahead of the package's copy. Nothing needs rebuilding
and nothing needs reinstalling.
To put the session screen back, remove the D-Bus shadow and the
`kglobalshortcutsrc` entries; both are package files, so an update will
restore them, which is the intended direction.